The Hypothesis

Can a marketer describe a retention flow in plain English and have it live in hours — with the system self-optimising from day one?

The Concept

The first five ideas in the stack are acquisition-heavy — getting people to click, land, and convert. But the most valuable marketing happens after the first conversion: onboarding, activation, upsell, churn prevention, win-back, advocacy. This experiment extends the stack into the full customer lifecycle. A marketer describes a flow in plain English — "when a new user hasn't completed onboarding after 48 hours, send a nudge focused on their most relevant feature" — and the AI builds the workflow: triggers, email content, timing, branching logic, and persona-aware messaging.

Lifecycle and retention engine

The hypothesis

Can a marketer describe a retention flow in plain English and have it live in hours — with the system self-optimising from day one?

The concept

The first five ideas in the stack are acquisition-heavy — getting people to click, land, and convert. But the most valuable marketing happens after the first conversion: onboarding, activation, upsell, churn prevention, win-back, advocacy. This experiment extends the stack into the full customer lifecycle. A marketer describes a flow in plain English — “when a new user hasn’t completed onboarding after 48 hours, send a nudge focused on their most relevant feature” — and the AI builds the workflow: triggers, email content, timing, branching logic, and persona-aware messaging.

How it works

1. Describe flow in plain English — intent, triggers, audience, outcome goal
2. AI builds workflow — triggers, content, timing, branching logic
3. Persona-aware messaging — different paths per cohort (budget buyer vs. enterprise evaluator)
4. Deploy and measure outcomes — activation rate, second purchase, churn rate, LTV — not just opens
5. Evolve continuously — system refines messaging, timing, and triggers based on retention outcomes

Describe the intent, set the outcome goal, and the system builds and optimises the lifecycle flow.

What it explores

• Can plain-English flow descriptions produce production-ready lifecycle automation?
• Does persona-aware pathing improve retention metrics vs. one-size-fits-all flows?
• Which lifecycle stage (onboarding, upsell, churn prevention) benefits most from AI optimisation?
• Can the system detect churn signals early enough to intervene effectively?
• Does AI-optimised timing outperform marketer-set timing for lifecycle messages?

What we found

• Onboarding completion improved 31% with AI-built flows vs. manual journeys
• Churn prediction flagged at-risk users an average of 12 days before cancellation
• Persona-aware pathing improved second-purchase rate by 19%
• AI-optimised send timing mattered more than content changes for onboarding nudges — the right message at the wrong time performed worse than an average message at the right time

Learnings

• Plain-English flow descriptions work for common patterns but need structured input for complex branching logic.

• Optimise timing before copy — send-time is the highest-leverage variable in lifecycle flows and the easiest to automate.

• Generic win-back messages (“we miss you”) underperform targeted re-engagement — churn intervention content needs to be as persona-aware as acquisition content.

• Persona-aware pathing pays off most at onboarding (different users need different activation paths) and least at win-back.

• Measure retention outcomes (LTV, second purchase) not engagement metrics (opens) — the two produce fundamentally different optimisation decisions.

Where it goes next

The timing finding is worth isolating. We’re designing a standalone experiment to test whether send-time optimisation alone — without any content changes — can meaningfully move retention metrics. If timing is truly the strongest variable, it changes how we think about lifecycle automation entirely.